194 F.D. Menalled et al. Agriculture, Ecosystems and Environment 77 2000 193–202
between crop field edges and crop field interiors Marino et al., 1997, among years Cardina et al.,
1996, or as a function of management practices Brust and House, 1988. At the within-field scale,
Marino et al. 1997 found the effect of post-dispersal seed predators to be patchy and not consistently re-
lated to the relative location of hedgerows. This led Marino et al. 1997 to suggest that future studies
of seed predation in agroecosystems should evaluate weed seed predation at a larger scale of analysis.
In agricultural ecosystems, most potential post- dis- persal seed predators such as small mammals Pol-
lard and Relton, 1970; Castrale, 1987, birds Lewis, 1969; Best, 1983 and insects Thomas et al., 1991,
1992 are found in non-crop habitats. Carabid bee- tles, which are important seed consumers in temperate
agroecosystems Johnson and Cameron, 1969; Best and Beegle, 1977; Lund and Turpin, 1977; Kjellsson,
1985; Brust and House, 1988; Manley, 1992 are also known to use non-crop habitats as over-wintering sites
Desender, 1982; Sotherton, 1984, 1985; Wallin, 1985; Thomas et al., 1991, 1992; Lys and Nentwig, 1992;
Lys et al., 1994; Zangger et al., 1994. Because of the linkage between seed predators and non-crop habi-
tat, it follows that the relative abundance of non-crop habitats in an agricultural landscape may have an ef-
fect on weed seed predation within crop fields. As such, it should be expected that predation on weed
seeds would be lower in simplified agricultural land- scapes than in complex agricultural landscapes. This
study assesses weed seed removal by seed predators in Michigan maize fields at the landscape scale. It com-
pares weed seed loss in complex agricultural land- scapes small crop fields embedded in a matrix of nu-
merous hedgerows and woodlots with that in sim- ple agricultural landscapes large crop fields embed-
ded in a matrix of widely scattered woodlots and hedgerows.
2. Methods
2.1. Landscape and hedgerow characterization The 3.2 by 12.9 km study region was located in
Onondaga Township, Ingham County, Michigan, USA 42
◦
25
′
30
′′
N, 84
◦
29
′
00
′′
W. This area was se- lected because it encompassed a gradient between
two typical agricultural landscapes of southern Michi- gan and was previously used by Marino and Landis
1996 and Menalled et al. 1999b to analyze the influence of landscape structure on insect parasitism
and parasitoid diversity. The southernmost 3.2 km
2
contains a highly heterogeneous mixture of crop and non-crop habitats hereafter ‘complex landscape’.
The northernmost 3.2 km
2
is a homogenous area with low crop diversity and few non-crop habitats here-
after ‘simple landscape’. The central area comprised a transitional landscape.
Landscape structure was characterized and quan- tified using black and white aerial photographs
1 : 2000 and digital land-use data. Photos were taken on 12 June 1988 and were scanned at 150 dpi
dots per inch and analyzed with ERDAS Earth Re- source Data Analysis System 7.5 ERDAS, Atlanta,
Georgia, USA. All agricultural fields within each landscape were identified and 30 fields from each
landscape type selected randomly for more intensive analysis. The following attributes of each selected
field were measured: area, perimeter, distance from the center of the field to the closest field edge, num-
ber of edge types per field, area to perimeter index, percentage of field perimeter comprising late succes-
sional habitats woodlots, wide hedgerows [10 m], narrow hedgerows [5–10 m], shrublands, and old
fields and percentage of field perimeter comprising early successional habitats herbaceous roadside and
crops and residential areas. Digital land use data from the Michigan Department of Natural Resources
Inventory System MIRIS were used to obtain an overall evaluation of land use patterns at a multi-field
scale. Using MIRIS data, the percentage of crop- land and deciduous habitats was determined for each
landscape type and contrasted between landscapes. Analysis of differences between landscape types
was conducted using t-tests with significance levels corrected using a sequential Bonferroni adjustment
Holm, 1979; Rice, 1988. Further details used to assess the structure of these landscapes can be found
in Marino and Landis 1996 and Menalled et al. 1999b.
2.2. Seed removal experiments Four conventional tillage corn Zea mays L. fields
in the complex and four conventional tillage maize
F.D. Menalled et al. Agriculture, Ecosystems and Environment 77 2000 193–202 195
fields in the simple landscape were chosen for the seed removal experiments. Fields were selected to repre-
sent typical cropping systems within each landscape. Because of the labor-intensive nature of this study it
was not possible to replicate simple and complex land- scapes. Also, crop rotation impeded the temporal repli-
cation of this study. Therefore, conclusions regarding landscape influences on weed seed removal are limited
to the areas under study.
Four agricultural weed species commonly found in Michigan were used to examine the effect of
agricultural landscape complexity on post-dispersal seed removal: crabgrass Digitaria sanguinalis L.
[Scop.], giant foxtail Setaria faberii Herrm., pig- weed Amaranthus retroflexus L., and velvetleaf
Abutilon theophrasti Medicus. Species used in this study were chosen to represent a wide range
of seed size seed size: D. sanguinalis = 2–2.2 mm, S. faberii = 2.5–3 mm, A. retroflexus = 1–1.2 mm, A.
theophrasti = 3–3.5 mm; Radford et al., 1968 and morphology Davis, 1993. Three treatments were
employed to quantify seed predation: 1 no ex- closure, allowing both vertebrate and invertebrates
to remove seeds 2 vertebrate exclosures, which prevented vertebrates from removing seeds but al-
lowed invertebrates to remove them, and 3 ver- tebrate + invertebrate exclosure, which prevented
both vertebrates and invertebrates from removing seeds and were used to determine unknown losses
of seeds. Vertebrate exclosures were constructed with cages of 1.25 sq. cm mesh rigid hardware cloth
34 cm × 34 cm × 7
cm length × width × height
sunk 3 cm into the soil. Vertebrate + invertebrate exclosures consisted of vertebrate exclosure cages
enclosing plastic rings 28 cm diameter, 5 cm high sunk 3 cm into the ground. Rings were painted with
Fluon
TM
, a slick material that prevents invertebrates from climbing the barrier and excludes them from
reaching the seeds placed within rings Mittelbach and Gross, 1984. Vertebrate + invertebrate exclo-
sures were utilized to estimate recovery efficiency and unknown sources of seed losses such as trans-
portation to and from the fields. For each treatment, 50 seeds of each species were placed on separate
11 cm × 14 cm × 0.5 cm
length × width × height waterproof pads 3M Metallic Finishing Pad. Seed
bank density in Michigan corn-soybean-wheat fields ranges between 1873 seeds m
− 2
and 5000 m
− 2
Renner et al., 1998. Fifty seeds were placed on each pad
3246 seeds m
− 2
to resemble natural occurring seed densities. Our laboratory observations indicated that
invertebrates walk freely on pads and that seeds were concealed in the rough surface of the pads, resem-
bling the situation observed for freshly shed seeds on soil. Pads were placed with one side flush with
the soil surface and were used to reduce seed losses from wind and facilitate recovery of uneaten seeds.
Each treatment was covered with a clear plastic roof to reduce seed losses from rain. In each field, one
repetition of each species-treatment combination was established at 27 m from the center of each one of
three randomly chosen field edges. This distance was chosen because this was the distance from the field
edge to the center in the smallest field. The order of the four species and three treatments within each
repetition was completely randomized with sample units 2 m apart. Thus, in each field, 36 sample units
four species, three treatments, and three repetitions were established using a total of 14,400 seeds in
each trial.
Field experiments were done twice during the period corresponding to the peak of abundance of
potential invertebrate seed predators such as carabid beetles Kirk, 1973. This period is usually associ-
ated with the period of natural weed seed production and dispersal. The first experiment was started on
3 September, 1996 and the second was started on 23 September, 1996. During the first trial, weather
conditions were dry with no heavy rains or winds, whereas it rained during the first 2 days of the second
trial. Seeds were left in the field for 1 week, recov- ered and the number of seeds remaining on all pads
was determined in the laboratory. Because seed coats and rodent fecal pellets were observed on a high
proportion of pads, seed removal was assumed to be primarily due to predation Schupp and Frost, 1989;
Myster and Pickett, 1993. Results were analyzed using a four factor landscape, field, species, and
treatment ANOVA model with fields nested within landscapes and proportion of seeds removed as the
dependent variable using Proc GLM, SAS software SAS Institute, 1996. To normalize the data and to
increase homoscedasticity, the proportion of seeds removed was modified using the arcsin Freeman and
Tukey transformation prior the ANOVA Sokal and Rohlf, 1995.
196 F.D. Menalled et al. Agriculture, Ecosystems and Environment 77 2000 193–202
3. Results
